Thesis_Defense

Report
A Comparison of Digital Elevation Models to
Accurately Predict Stream Channels
Papillion
Creek
Watershed
Spencer Trowbridge
Presentation Outline
Research Objectives
Methodology
Results
Summary
Research Objective: Determine the resolution of a digital
elevation model that best predicts stream channel
locations
Hypothesis: A finer resolution DEM does not affect its ability
to accurately predict a stream channel’s location
Significance of Research
Offers a new way to evaluate a DEM’s ability to extract
stream channels
LiDAR Dataset has not been evaluated for this application
Determination of Error
Distance of
derived
streams to
traced
banks was
classified as
error
National Elevation Dataset (NED)
National Elevation Dataset (NED)
Based on USGS topographic maps
Continually updated with most recent and accurate
methods available
– April 2014 using LiDAR data
Shuttle Radar Topography Mission (SRTM)
Shuttle Radar Topography Mission (SRTM)
Worldwide coverage from 2000
A variety of resolutions available (90 meters for this study)
Light Detection and Ranging (LiDAR)
Data supplied as raw LAS files
High precision elevation data captured and stored as point
Flown in 2010 – Same time as aerial photographs used for
tracing stream channels
Study Area
• Hilly
topography
• Urbanized
areas with
channelized
streams
Named Streams
Big Papillion Creek
North Branch
South Papillion Creek
Mud Creek
West Papillion Creek
Leach Branch
Little Papillion Creek
Hell Creek
Papillion Creek
Falls Branch
Thomas Creek
East Fork
Cole Creek
Copper Creek
Northwest Branch
Butter Flat Creek
Walnut Creek
Boxelder Creek
Southwest Branch
Boston Branch
Richter Branch
Big Elk Creek
Methodology
•
Data Gathering
•
SRTM, NED, NHD, LiDAR, Orthophotographs
•
LiDAR DEM Creation
•
Stream Extraction
•
ArcGIS Hydrology tools:
• Fill
• Flow Direction
• Flow Accumulation
• Stream to Feature
•
Stream Tracing
•
Digital Shoreline Analysis System (DSAS)
•
Baseline Drawing, Transect Casting
Raster Data
SRTM DEM
NED DEM
1:800,000 Clipped Data
Mosaic of two rasters
1:800,000
Raster Data Resolution Differences
SRTM 90 meters per pixel
1:25,000
NED 30 meters per pixel
1:25,000
Channels clearly visible
Stream Locations
SRTM 90 meters per pixel
NED 30 meters per pixel
Little Papillion
Creek
Big Papillion Creek
1:25,000
Little Papillion
Creek
Big Papillion Creek
1:25,000
1:25,000
Visual Comparison of Stream Locations
Little Papillion
Creek
Big Papillion Creek
1:25,000
LiDAR DEM Creation
Suggested workflow for raster creation from LiDAR data
according to ESRI help pages. http://resources.arcgis.com
1. Create LAS Dataset to hold the data
2. Point File
3. LAS to Multipoint
4. Create Terrain
5. Terrain to Raster
LiDAR Point File Tool
Determination of average point spacing - Should be
available in the metadata of the LAS files. If missing, run
Point File tool
LiDAR DEM
Result: 4.34 meters
LAS to Multipoint tool asks for
average point spacing
4.34 meter cell size
1:50,000
LAS to Multipoint
Creates large feature class with elevation point data
Create Terrain Tool
Creates triangulated surface from points from LAS to Multipoint
1:1,000
1:15,000
Terrain to Raster
Final LiDAR Raster
SRTM
90 meters
NED
30 meters
1:22,000
4.34 meter cell size
Stream Extraction
ArcGIS hydrology tools used
Fill
Flow Direction
Flow Accumulation
Raster Calculator
Stream to
Feature
Creates vectors connecting
cells from the flow
accumulation tool
Stream to Feature
LiDAR Stream to Feature after Raster Calculator tool
1:12,000
1:300,000
Lower order streams eventually
edited out manually
NHD Vector Data
NHD stream vectors
downloaded and
Papillion Creek
Watershed after
removal of
unneeded streams
used as verification
1:365,000
1:250,000
SRTM DEM
1:40,000
1:250,00
NED DEM
1:40,000
LiDAR DEM
1:40,000
All Streams
All Streams
LiDAR
NED
SRTM
NHD
1:40,000
1:300,000
DSAS
ArcGIS extension for collecting shoreline distance
measurements
• Manually drawn baselines
• Casting transects
• Raw data collection
• Quality control
Manually Drawn Baselines
Baselines
follow the
general
direction of
the stream
channel
Casting Transects
Transects cast
orthogonally from
the baseline
Extend far enough
to Intersect all
extracted streams
Stream Tracing
Once transects
are cast, one
knows where
they will intersect
with the traced
stream banks
Stream and
transect
intersection point
Stream Tracing
Traced streams
edited where
they intersect
transects
Generalized
Stream Tracing
Transect
Stream
Bank
Precise
Stream Tracing
at transect
crossing
Raw Data Collection
Using the
appropriate
streams as input,
DSAS automatically
calculates
intersection
distances
Results
Error Distance Calculations
Correct Stream Placement Determination
Distribution Analysis of Datasets: Normality for ANOVA
ANOVA
Student-Newman-Keuls Test Results
Determination of Error
Distance of
derived streams to
traced banks was
classified as error.
These values were
manually
calculated within
a spreadsheet and
used for analysis.
Determination if Streams Predicted Accurately
Within a spreadsheet,
samples determined if
predicted accurately
(within the traced
stream)
If a sample was found
to be within the traced
banks, a distance
value of zero was
assigned
Box-Cox Transformation
NED
LiDAR
NHD
SRTM
Normal Distributions
Bimodal Distribution
ANOVA
Student-Newman-Keuls
Summary
• Resolution of DEMs affects their ability
to delineate stream channels
• The finest resolution DEM is not the
best at delineating stream channels
Conclusions
•
A method was created to evaluate a DEM
•
LiDAR data can be processed (and manipulated)
in a variety of ways
• It may not be necessary to have high resolution
LiDAR data depending on the application
•
LiDAR is not meant for huge areas such as this
watershed
Questions
All Streams
LiDAR
NED
SRTM
NHD
1:40,000
1:300,000

similar documents